The next major step in mobile radio communications

22

Initial field performance measurements of LTE

Researchers have measured the performance of LTE in the field using different drive routes and radio channel environments. Among other things, they measured different bandwidths, different antenna configurations and layer-1, UDP and TCP throughput using two kinds of terminals.
Jona s K a r l s s on M at h i a s r i bac K

The 3GPP Long Term Evolution (LTE) standard has made great strides during the past year. Measurements from the LTE SAE Trial Initiative (LSTI) have moved ahead as scheduled and, as expected, shown good results.1 In January 2008, 3GPP confirmed that the technology specifications for the LTE Terrestrial Radio Access Network had been approved and were under Change Control, which means they will be included in the forthcoming 3GPP Release 8.2
At Mobile World Congress 2008, in Barcelona, major CDMA and WCDMA operators, such as NTT DoCoMo,

Verizon Wireless, Vodafone and China Mobile, confirmed their commitment to LTE as the next-generation mobile network. And wireless industry leaders have reached an agreement concerning a framework for licensing technology IPR. This article describes some of the field-performance measurements made using different drive routes and radio channel environments while traveling at velocities of up to 120km/h and at distances of up to 4km. The trials include measurements of both 10 MHz and 20 MHz bandwidths and of different antenna configurations up to 44 MIMO. The article also presents the layer-1, UDP and TCP throughput using two different kinds of terminals: a large terminal (the test bed, built on DSP technology) and a prototype handheld terminal. Please note that the performance figures are

taken from the test-bed results; the actual performance of commercial products might differ somewhat. LTE The LTE standard is being developed in two 3GPP work items: The LTE work item targets the evolution of the radio network; and the SAE (System Architecture Evolution) work item targets the evolution of the packet core network. In contrast to previous standards, the LTE standard solely specifies a packet-switched (PS) domain for LTE and SAE. Output from the work items is embodied in the Evolved UTRAN (E-UTRAN) and the Evolved Packet Core (EPC) standards. Together, they form the Evolved Packet System (Figure 1). The E-UTRAN standard employs orthogonal frequency-division multiplexing (OFDM) technology for downlink operation and singlecarrier frequency-division multiple access (SC-FDMA) technology for uplink operation. This approach yields excellent flexibility in terms of deploying and allocating spectrum from 1.4 MHz up to 20 MHz. In addition, the E-UTRAN standard supports FDD (frequency-division duplex) as well as TDD (time-division duplex) modes of operation. The basic LTE downlink physical resource can be seen as a timefrequency grid (Figure 2), where each resource element corresponds to one OFDM subcarrier during one OFDM symbol interval. Downlink subcarrier spacing has been set at 15kHz. LTE uplink transmission is based on SC-FDMA with dynamic bandwidth. The single-carrier property of the uplink, which limits transmission to a contin-

BOX A

Terms and abbreviations

Third Generation Partnership Project Additive white Gaussian noise Cumulative distribution function Digital signal processor Ericsson Mobile Platforms Evolved packet core Evolved packet system Evolved UTRAN Extended vehicular A Frequency-division duplex Global positioning system Intellectual property rights International Telecommunication Union Line of sight LTE SAE Trial Initiative Long Term Evolution Multiple input, multiple output Next-generation mobile network Orthogonal frequency-division multiplexing PB3 Pedestrian B 3km/h PS Packet switched SAE System Architecture Evolution SC-FDMA Single-carrier frequency-division multiple access SNR Signal-to-noise ratio S-PARC Selective per-antenna rate control TCP Transmission control protocol TDD Time-division duplex UDP User datagram protocol UMTS Universal mobile telecommunications system UTRAN UMTS terrestrial radio access network XPD Cross-polar discrimination NGMN OFDM

3GPP AWGN CDF DSP EMP EPC EPS E-UTRAN EVA FDD GPS IPR ITU LOS LSTI LTE MIMO

E r i c s s o n r E v i E w 3 2008

23

Simplified architecture of the Evolved Packet System.

FIgurE 1

Internet, operator service, etc.

LTE test bed is a one-sector, one-cell system designed for early testing and proofof-concept. Some updates have been made to the hardware since 2006, but it is essentially the same flexible, highperformance platform described in LTE test bed.3 It supports
a variety of configurations, antenna setups, and system bandwidths; and measurements of layer-1, layer-2 and application-level performance.

BOX B

3gPP LTE
3GPP LTE (Long Term Evolution) is the name given to a project to improve the UMTS mobile phone standard to cope with future technology evolutions. Goals include improving spectral efficiency, lowering costs, improving services, making use of new spectrum and refarmed spectrum opportunities, and better integration with other open standards.

to perform modem use cases with a laptop; to demonstrate multimedia telephony (for example, VoIP); to demonstrate video telephony; to perform file transfers; to stream video applications; and to test multiple terminals simultaneously.

EPS

EPC

E-UTRAN

E-UTRA Terminal

uous frequency block for each terminal, greatly improves power efficiency in terminals. One important aspect of LTE is its support of various multi-antenna schemes. In addition to receive diversity, both at the base station and at terminals, LTE downlink transmission supports openloop transmit diversity, two types of spatial multiplexing, and the scheduling of multiple terminals on the same time and frequency resource a technique called multi-user MIMO (multiple-input multiple-output). Spatial multiplexing (MIMO) makes it possible to transmit parallel streams of data, which substantially increases the peak data rate. In theory, the peak rate of a system with a 22 antenna configuration (two transmit and two receive antennas) using two-stream MIMO should be double that of a 12 system (one transmit and two receive antennas). Likewise, the theoretical peak rate of a 44 system with four-stream MIMO should be four times that of a 12 system. Ericsson LTE test bed The hardware architecture of Ericssons

The baseband software is based on a common platform, but different versions of the code are used for different hardware configurations in order to match the current antenna configuration. Test bed development has taken place in parallel with the 3GPP standardization process. Therefore, the LTE test bed does not exactly mirror the most recent 3GPP LTE specification. The MIMO scheme in the test bed, for example, employs selective per-antenna rate control (S-PARC), which means streams can be individually encoded and modulated and a single data stream is always mapped to a single transmit antenna.4 In addition to measurements made using the LTE test-bed hardware, this article includes measurements performed with a prototype handheld LTE device developed by Ericsson Mobile Platforms (EMP) and shown at Mobile World Congress 2008 in Barcelona (Figure 3). The terminal, which is compatible with the base station, has been used
FIgurE 2

Experimental The results were obtained using the test bed in the laboratory and in extensive outdoor field tests. Fading emulators and additive white Gaussian noise (AWGN) sources were used in the lab to measure the performance of different channel profiles at different signal-to-noise ratios (SNR). For the field trials, an LTE test site was installed on the roof at Ericssons headquarters in Stockholm (Figure 4). Commercially available antennas for the 2.6 GHz band were used at the base station and in the terminal. To evaluate the use of spatially separated and dualpolarized antennas, the testers used different antenna setups and benchmarked them against each other for different MIMO configurations. The test-bed terminal was installed in a measurement van with antennas mounted on the roof (Figure 5). A GPS receiver was also used to make it possible to associate logged performance with the position of the terminal. The field tests were performed in two sectors that represent different instal-

The LTE downlink physical resource grid.

Subcarrier spacing = 15 kHz

One resource element Tim

One OFDM symbol

Frequency

E r i c s s o n r E v i E w 3 2008

The next major step in mobile radio communications

24

FIgurE 3

Berta Ericssons prototype handheld LTE device.

lation scenarios. The first sector is quite sparsely settled and dominated by low-rise buildings (compared to the site installation height) line-of-sight (LOS) conditions were thus fairly common. The second sector is more dense and building heights are comparable to the site installation LOS conditions were thus quite rare. The majority of the field tests were performed within a radius of 1km from the test site. The tests also included distances of more than 4km. During most parts of the field tests, the speed of the terminal was between 5 and 40km/h. Some of the tests were conducted while driving on a freeway at speeds of more than 100km/h. Measurements Lab results Figure 6 contrasts the layer-1 throughput with SNR of different antenna configurations. The measurements, made with 10 MHz bandwidth and the Extended Vehicular A (EVA) 3km/h channel model5, were complemented with an emulated cross-polar discrimination (XPD) of 10dB, which simulates the use of dual-polarized antennas. As anticipated, the 14 setup outperforms the 12 setup, especially when SNR is low. Compared to the 12 setup, the 22 setup yields gains of approximately 10 to 15dB and up. By contrast, the 24 setup, yields multi-stream gains of around 5dB. The gain with four streams compared to two is seen from 20dB and up. With this specific channel model, the peak throughput of the 44 setup yields a gain of approximately 150 percent compared to the 12 setup. Other channel models yield different gains Figure 7, for example, shows the gain that can be obtained from using an ideal 44 AWGN MIMO channel. To evaluate application level performance, measurements were taken using UDP and TCP traffic over a 22 20 MHz MIMO Pedestrian B 3km/h (PB3) channel.6 As can be seen in Figure 8, the difference in performance between UDP and TCP is very small. UDP is a best-effort transmission technique without acknowledgements from the terminal to the base station. It is thus insensitive to uplink performance and IP-related issues, such as system buffer sizes. TCP, on the other hand, transmits

FIgurE 4

Installation of the base station antenna.

E r i c s s o n r E v i E w 3 2008

25

TABLE 1

Example of relative gain compared to 12 antenna setup.

Throughput gain relative 1 x 2 setup 10th percentile 1x4 2x2 2x4 4x4 + 15 % + 27 % + 86 % + 104 % Mean + 12 % + 42 % + 87 % + 108 % 90th percentile +1% + 55 % + 91 % + 100 %

acknowledgements in the uplink. The results indicate that the test system has a well-tuned and reliable uplink. Field results In good radio conditions, the system achieved the theoretical maximum throughput over the downlink of a 22 system using link adaptation and 20 MHz bandwidth: 170Mbps. Using four transmit streams (the maximum number supported in the LTE standard), four receive antennas (44 MIMO) and 10 MHz bandwidth, the measured peak rates exceeded 130Mbps. This translates into approximately 260Mbps, given the maximum bandwidth of 20 MHz, which is more than three times the peak throughput that can be achieved from a basic 12 setup. Figure 9 shows the cumulative distribution functions (CDF) for different antenna configurations along a drive route where vehicle velocity was between 5 and 40km/h. The measurements show that adding more transmit and receive antennas improves performance. Table 1 compares the relative gains to a 12 setup along a given drive route. The MIMO-related gains are strongly dependent on radio channel conditions and should therefore solely be seen as an example of what can be achieved. From the measurements, including those derived from different drive routes, it is seen that adding a second stream at the base station (22) can increase the mean throughput by approximately 15 percent to 40 percent and more or less double the peak throughput. Adding two more receive antennas but keeping a single transmit antenna at the base station improves

the mean throughput by 10 percent to 20 percent compared to the 12 setup. Combining a second stream at the base station with four receive antennas (24) significantly increases the mean throughput (50 percent to 70 percent) compared to a 12 setup. Large multistream gains were seen in both sectors using dual-polarized as well as spatially separated antennas. The addition of more transmit streams (up to four) increases peak throughput by approximately 60 percent, whereas the mean throughput increased by 8 percent to 25 percent compared to the 24 setup. In conclusion, besides increasing peak data rates, we see that multi-stream transmission improves the mean and 10th percentile data rates.
FIgurE 5

Figures 10 and 11show performance along a drive route using 10 MHz bandwidth for 44 and 22 MIMO setups. Good throughput was achieved along the entire route. In some areas the peak rates were very high. The 44 setup performed better than the 22 setup except in line-of-sight conditions where performance was similar. The lab measurements (Figure 11) assessed the TCP and layer-1 bit rates available for applications using 20MHz bandwidth. The difference between layer-1 bit rates and corresponding TCP bit rates can be attributed to protocol overhead. The layer-1 bit rate (no protocol overhead) is available to higher-level applications. The TCP bit rate was more than 40Mbps at least 50 percent of the time and more than 100Mbps at least 10 percent of the time. The impact of velocity on performance was evaluated on a freeway. Given good radio conditions, a bit rate of more than 100Mbps was obtained while traveling at more than 100km/h and using 20 MHz bandwidth. Under the same conditions, but at a distance of more than 4km from the base station, throughput was 40Mbps or greater. Field measurements using the prototype handheld terminal were conducted in Nuremberg, Germany

Test measurement van.

E r i c s s o n r E v i E w 3 2008

The next major step in mobile radio communications

26

FIgurE 6 Layer-1 throughput measured at 10 MHz bandwidth using Extended Vehicular A (EVA) 3km/h channel model.

FIgurE 7 Layer-1 throughput measured at 10 MHz bandwidth using full rank AWgN channel model.

Throughput (Mbps) 80 70 60 50 40 30 4 x 4 MIMO 2 x 4 MIMO 1 x 4 receive diversity 2 x 2 MIMO 1 x 2 receive diversity 10 MHz bandwidth

Throughput (Mbps) 120

100

4 x 4 MIMO 2 x 4 MIMO 1 x 4 receive diversity 2 x 2 MIMO 1 x 2 receive diversity

10 MHz bandwidth

80

60

40 20 10 0 0 5 10 15 20 25 30 35 SNR (dB) 20

10

15

20

25

30

35 SNR (dB)

FIgurE 8 Performance of uDP and FTP at 20 MHz bandwidth using 22 setup with Pedestrian B 3km/h (PB3) channel model.

FIgurE 9 Cumulative distribution functions (CDF) for different antenna configurations measured with 10 MHz bandwidth and dual-polarized antennas.
CDF 1.0

Throughput (Mbps) 90 80 70 60 50 UDP, 64QAM, R=0.5 FTP, 64QAM, R=0.5 UDP, 16QAM, R=0.5 FTP, 16QAM, R=0.5 UDP, QPSK, R=0.5 FTP, QPSK, R=0.5 20 MHz bandwidth

0.9 0.8 0.7 0.6 0.5 10 MHz bandwidth 1x2 1x4 2x2 2x4 4x4

40 30 20 10 0

0.4 0.3 0.2 0.1 0 5 10 15 20 25 30 35 SNR (dB) 0 0 20 40 60 80 100 120 Throughput (Mbps)

E r i c s s o n r E v i E w 3 2008

27

(Figure 12). Commercially available antennas were installed on the fourth floor at Ericssons premises. The frequency band was 2.6 GHz; bandwidth was 20 MHz. In a static measurement made close to the base station (Point A), the maximum layer-1 bit rate over the downlink was 25Mbps. This is also the maximum throughput of the prototype terminal in this configuration when connected via cable. UDP throughput over the downlink was close to 23Mbps. As the terminal moved away from the base station (Point B), the maximum layer-1 data rate remained steady at 25Mbps, whereas UDP throughput fell slightly to just above 20Mbps. Driving with the terminal at close to 30km/h, the layer-1 data throughput over the downlink was around 22Mbps; the UDP data rate was around 18Mbps. To further demonstrate the strength of the prototype, three additional terminals were placed in radio conditions that varied from high to considerably low SNR. Despite a variance in SNR of more than 20dB, all three terminals reached a throughput of about 14Mbps. Keeping in mind that the terminal is still an early prototype, these performance figures are most promising. Summary and future perspectives Measurements made in the lab and extensive field trials show that LTE performs well in both the physical and application layers. Multistream MIMO yields good gains in realistic environments, improving peak rates as well as the mean and 10th percentile throughput. A small, prototype handheld LTE terminal has been used, among other things, to stream video from an LTE test-bed base station and to illustrate that LTE is quickly becoming a mature technology. Work is also under way to standardize the next releases of 3GPP to ensure that future end-user and operator requirements are met. The International Telecommunication Union (ITU) is defining the requirements for IMT Advanced, sometimes also referred to as 4G. Some of the requirements being discussed are bit rates of up to 1Gbps in the downlink, more than 500Mbps in the uplink, and a scalable bandwidth of up to 100 MHz. Other requirements

FIgurE 10 Left: Throughput relative to location for a 4x4 MIMO setup using 10 MHz bandwidth and dual-polarized antennas. right: Throughput relative to location for a 2x2 MIMO setup using 10 MHz bandwidth and dual-polarized antennas. Base station located at X in both plots.

FIgurE 11 TCP and layer-1 throughput for a 2x2 setup with 20 MHz bandwidth.

CDF 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 20 40 60 80 100 120 140 160 Throughput (Mbps) TCP Layer 1 20 MHz bandwidth

E r i c s s o n r E v i E w 3 2008

The next major step in mobile radio communications

28

FIgurE 12 Field testing the prototype handheld LTE terminal in Nuremberg, germany.

Jonas Karlsson
received an M.Sc. from Linkping University (Sweden), a Licentiate degree from the Royal Institute of Technology (Sweden) and a Ph.D. from the University of Tokyo (Japan), all in electrical engineering. Since 1993 he has worked for Ericsson Research in Sweden and Japan. The focus of his research has been on advanced antennas and interference cancellation. In 2007 he moved to the LTE System Management department, where he is a Senior Specialist in multiple antenna systems.

being discussed have already been met by LTE. The 3GPP has begun working on LTE Advanced, which is scheduled for inclusion in 3GPP Release 10. LTE Advanced should fulfill, but not be limited to, the IMT Advanced requirements. The 3GPP view is that LTE Advanced should
aim for NGMN (next-generation mobile network) system performance. The NGMN intends to complement and support the work within standardization bodies by providing a coherent view of what the operator community will require in the decade beyond 2010; fulfill 3GPP compatibility requirements that is, it should allow LTE Release 8 terminals in LTE Release 10 networks and vice versa; and support smooth migration from LTE Release 8.

Mathias riback
joined Ericsson in 2004 to work as a research engineer at Access Technologies and Signal Processing, Ericsson Research, Stockholm. The focus of his work has been on radio propagation research and test-bed development (HSPA and LTE) with emphasis on multi-antenna aspects. Mathias is currently responsible for the LTE test bed activities at Ericsson Research. He holds an M.Sc. in electrical engineering from the Royal Institute of Technology (Sweden).

Acknowledgements
Werner Anzill, Robert Jansen, Hanna Maurer-Sibley and David Wisell

In addition, the 3GPP is studying extended multi-antenna deployments as well as ways of lowering terminal and network power consumption.

references
1. www.lstiforum.org 2. The Third Generation Partnership Project, www.3gpp.org. 3. Johansson, B. and Sundin, T.: LTE test bed. Ericsson Review, vol. 84(2007):1 pp. 913 4. Grant, S., Molnar, K. and Krasny, L,: System-level performance gains of per-antenna-rate-control (S-PARC), VTC Spring 2005, vol. 3, pp. 16961700 5. 3GPP spec. 36.101, ver. 8.1.0, Annex B 6. 3GPP spec. 25.101, ver. 8.2.0, Annex B

E r i c s s o n r E v i E w 3 2008